📄 中文摘要
研究社区在星际争霸 II 的完整游戏与其迷你游戏之间缺乏一个中间选择。完整游戏的庞大状态-动作空间导致奖励信号稀疏且嘈杂，而在迷你游戏中，简单的智能体则表现出饱和的性能。这种复杂性差距妨碍了稳定的课程设计，并阻止研究人员在现实计算预算下对现代强化学习算法进行实验。为填补这一空白，提出了双桥地图套件，这是一个开源基准系列的首个条目，专门设计为介于这两者之间的中间基准。通过禁用经济机制，如资源收集、基地建设和视野遮蔽，该环境隔离了两个核心战术。
规模化策略，而非计算：一个独立的开源星际争霸 II 基准测试，促进可及的强化学习研究
📄 English Summary
Scaling Strategy, Not Compute: A Stand-Alone, Open-Source StarCraft II Benchmark for Accessible Reinforcement Learning Research
The research community lacks a middle ground between the full game of StarCraft II and its mini-games. The sprawling state-action space of the full game results in sparse and noisy reward signals, while simple agents saturate performance in mini-games. This complexity gap hinders steady curriculum design and prevents researchers from experimenting with modern Reinforcement Learning algorithms in RTS environments under realistic compute budgets. To address this gap, the Two-Bridge Map Suite is introduced as the first entry in an open-source benchmark series purposely engineered to serve as an intermediate benchmark. By disabling economy mechanics such as resource collection, base building, and fog-of-war, the environment isolates two core tactical elements.